RigidBody (Node -> Node2D -> PhysicsBody2D -> RigidBody2D) Sprite (Node -> Node2D -> Sprite) CollisionShape (Node -> Node2D -> CollisionShape) So, at first it seems that the approach in Godot is more wasteful, but is it really? Node is lightweight, similar to a component. Node2D contains the 2D transform, similar to Transform component in ECS. Three nodes are required whereas one component suffices in ECS. This seems wasteful, but is it really? In the entity, the collider and the sprite will most likely not be used centred and will still require offset and rotation properties, so in the end not much changes.